when heirs become major shareholders 20140606 - uni … · 2019. 9. 29. · munich personal repec...
TRANSCRIPT
-
Munich Personal RePEc Archive
When Heirs Become Major Shareholders:
Evidence on Tunneling and Succession
through Related-Party Transactions
Hwang, Sunwoo and Kim, Woochan
February 2014
Online at https://mpra.ub.uni-muenchen.de/56487/
MPRA Paper No. 56487, posted 12 Jun 2014 18:14 UTC
-
- 1 -
When Heirs Become Major Shareholders*
Evidence on Tunneling and Succession through Related-Party Transactions
Sunwoo Hwang† and Woochan Kim
‡
This Draft: February, 2014
Abstract
In family firms, the succession of controlling equity stake to next generation is an issue of paramount
importance. This, however, can be a major challenge in the presence of heavy inheritance or gift tax
burden (high tax rate and absence of tax-saving vehicles, such as trusts or foundations) and in the absence
of dual-class equity. Such regulatory environment may lead families to seek alternative ways of succession.
As for families controlling business groups, one way of doing so is making use of related-party
transactions among member firms. By favoring firms where the heir holds significant equity stake, the
family can tunnel corporate resources to the heir. Eventually, the firm can grow large enough to acquire
controlling equity stakes in other firms within the group. In this paper, we investigate this possibility using
Korean chaebol firms during a sample period of 2000-2009. We identify firms where heirs become a
major shareholder (treatment group) and compare them against their year-industry-size-matched firms
(control group) before and after the ownership change. Difference-in-differences test with firm fixed
effects reveal that treatment group firms experience greater related-party transactions, benefit from them
in terms of earnings, pay out more dividends, and become more important in controlling other firms in the
group.
JEL classification: G30, G32, G34
Keywords: family firm, business group, chaebol, succession, related-party transactions, control-enhancement
* We thank workshop participants at Korea University Business School, Seoul National University
Business School, and Sungkyunkwan University Business School for their comments. We also thank KDI
School of Public Policy and Management and Asian Institute of Corporate Governance (AICG) for
financial support. † Kenan-Flagler Business School, The University of North Carolina at Chapel Hill, McColl Building,
Chapel Hill, NC 27599, USA, E-mail: [email protected] ‡ Corresponding author; Associate Professor of Finance, Korea University Business School, Anam-Dong,
Seongbuk-Gu, Seoul 136-701, Republic of Korea, E-mail: [email protected]
-
- 2 -
1. Introduction
In family firms, the succession of controlling equity stake to next generation is an issue of
paramount importance. A successful succession allows the family to perverse its control for
another generation. Depending upon who succeeds the equity stake, it will also greatly influence
the decision on who will be the next CEO. One can say that ‘managerial’ succession is of
secondary importance compared to ‘ownership’ succession.
In the existing finance literature, there is a growing body of research on family firm
performance (McConaughy et al., 1998; Anderson and Reeb, 2003; Maury, 2006; Villalonga and
Amit, 2006; Miller et al., 2007; Andres, 2008), the management succession of family firms
(Smith and Amoako-Adu, 1999; Bennedson et al., 2007; Cucculelli and Micucci, 2008) and the
control-enhancing mechanisms family firms use (La Porta, López de Silanes, and Shleifer, 1999;
Claessens, Djankov, and Lang, 2000; Nenova, 2001; Faccio and Lang, 2000; Villalonga and Amit,
2009; Gompers, Ishii, and Metrick, 2010). But surprisingly, no paper exists on the succession of
family ownership. This paper aims to make a small step in filling this gap .
Handing over controlling equity stake from one generation to another generally faces two
challenges. One is the risk of dilution and the other is the risk of taxation. If a family firm
repeatedly relies on external equity financing, the equity stake later generations inherit may not
be large enough to warrant control over the firm. In certain jurisdictions, this challenge is
resolved with the use of dual-class equity or voting agreements (Villalonga and Amit, 2009).
Descendants that hold shares with multiple voting rights or that entered contract with
shareholders ceding their voting power can be free from the risk of dilution.
The risk of taxation is another major challenge. Although some jurisdictions have abolished
inheritance tax, many still retain it.1 In the U.S. the tax rate is as high as 35%. Also, even if
abolished, capital gains tax may still apply upon the transfer of wealth. Gift tax also has been
abolished in recent years, and even if it exists, many deductions and exemptions apply.
Nevertheless, there are nontrivial number of jurisdictions that still retain it. In certain
jurisdictions, this challenge of taxation is resolved with the use of trusts or private foundations
1 Some jurisdictions use the term estate tax instead of inheritance tax.
-
- 3 -
that receive shares as donation (Villalonga and Amit, 2009). As charitable entities, they are
exempt from taxation, but still governed by family members who serve as trustees or board of
directors.
What if dual-class equity is prohibited by law, voting agreement counterparties are hard to
find, and trusts or private foundations are heavily regulated? Founding family would have a
strong incentive to seek alternative ways of handing over controlling equity stake to their
descendants. One alternative is forming a business group and letting the heir own a controlling
equity stake in a firm or multiple firms that control others. The business group can have a
pyramidal structure, but this is not necessary. Cross- or circular-shareholdings can also serve the
same purpose. But, there is a critical problem with this scheme. The heir may not have enough
wealth in the beginning to acquire controlling equity stake in the firm that controls others.
There are two possible solutions to this. One is making the holding company or the de facto
holding company (in case the group does not have a pyramidal structure) issue new shares
privately to the heir at a heavily discounted price, so that the heir can acquire controlling equity
stake in the holding company with low financial burden.2 This, however, is not possible for
publicly-traded companies, where preemptive rights of existing shareholders are typically well
protected. Even for privately-traded ones, tax implications will prevent the use of such scheme.
An alternative solution is tunneling corporate wealth from one firm to another through
related-party transactions. That is, setting up a privately-traded firm, where the heir is a major
shareholder, and instructing other firms to purchase goods and services from that firm. Increased
sales and earnings of this firm will increase its asset size. Eventually, the firm can grow large
enough to acquire controlling equity stakes in other firms within the group. The firm may also
pay out dividends so that the heir can directly acquire shares in other firms. Thus, family
ownership can be in the hands of the heir without paying any gift or inheritance tax to the
government.
This possibility implies that any serious research on the succession of family ownership must
also consider changes in intra-group ownership structure and related-party transactions among
2 Same purpose can be served with a convertible bond or a bond with warrant with heavily discounted
conversion ratio or exercise price.
-
- 4 -
member firms. In this paper, we attempt to do this by studying a country – Korea – that perfectly
matches the setting we initially considered. That is, a country where dual-class equity is
prohibited by law, voting agreement counterparties are hard to find, and trusts or private
foundations are heavily regulated.
Anecdotal evidence of family ownership succession through tunneling abounds among
Korean chaebols – family-controlled business groups in Korea. An exemplary case is Hanwha
S&C, an integrated IT service firm of Hanwha group (also see Figure 1). Originally, it was
wholly owned by Kim Seung-youn (33.3%), the group chairman of Hanwha group, and Hanwha
Corp (67.7%). But, by 2007, the shares of Hanwha S&C was sold to the chairman’s three sons,
each owning 50%, 25%, and 25%.3 Since then, Hanwha S&C’s sales to member firms soared
from 117 billion Korean won (approximately, 117 million US dollars) in 2007 to 319 billion
Korean won in 2010. Its earnings (EBIT) also jumped from 11 billion Korean won in 2007 to 24
billion Korean won in 2010. This improved financial strength enabled Hanwha S&C to acquire
shares in other member firms. As of 2012, it holds shares of Hancomm (70%), Hanwha
Corporation (2.2%), Hanwha Total Energy (100%), Hanwha General Insurance (0.37%), Yeosu
Cogeneration System (100%), Hanwha Solar Energy (20%), and Human Power (100%). Prior to
2007, Hancomm was the only firm, in which Hanwha S&C held shares.
To test our predictions using Korean chaebols during the sample period of 2000-2009, we
identify firms where heirs become a major shareholder (treatment group) and compare them
against their year-industry-size-matched firms (control group) before and after the ownership
change. Difference-in-differences test with firm fixed effects reveal a number of results consistent
with our predictions. First, related-party transactions increase in treatment group firms after the
3 In May 2010, the shareholders of Hanwha Corp. filed a derivative suit against the directors of Hanwha
Corp. for selling Hanwha S&C shares below the DCF value. In this civil charge, the shareholders asked
for a compensation of 45 billion Korean won (approximately, 45 million US dollars). In October 2013,
Seoul Central District Court ordered the directors to pay back to the company 8.9 billion Korean won,
which is well below the originally estimated damage. At the time of this writing, the case is at the
appellate court. In a separate criminal case (embezzlement), Chairman Kim was sentenced a three-year
prison with a five-year suspension (finalized in February 2014). But, he was acquitted from the charge of
selling Hanwha S&C shares below the DCF value. These results indicate how difficult it is to prevent
tunneling with ex post legal remedies.
-
- 5 -
treatment. Second, earnings increase with related-party transactions in treatment group firms after
the treatment. Third, dividend payout increase with related-party transactions in treatment group
firms after the treatment. Fourth, importance in group control increases with related-party
transactions in treatment group firms after the treatment. We measure the importance in group
control by marginal contribution to group control used in Kim, Lim, and Sung (2007). Further
analyses reveal that our results are driven by related-party sales to member firms rather than
related-party purchases from them.
We also conduct a number of falsification tests. First, we run similar difference-in-differences
regressions using treatment group firms where ‘non-heirs’ become a major shareholder. We do
not find any increase in related-party transactions in these firms after the treatment. Second, we
run similar difference-in-differences regressions where counterparties of the original treatment
group firms (e.g., firms where heir become a major shareholder) are used as our new treatment
group. Again, we do not find any increase in earnings, dividend payouts, or control over other
firms in these new treatment group firms after the treatment.
Our paper contributes to the existing literature in three ways. First, to the best of our
knowledge, this is the first paper on ‘family ownership’ succession, which we believe is an issue
of paramount importance in family firms, but its research virtually missing in the existing finance
literature. As mentioned earlier, the main focus of existing literature is on ‘managerial’ succession.
Second, we contribute to family firm performance studies, which became popular since
Anderson and Reeb (2003). We contribute to this area of study by highlighting the importance of
related-party transactions when assessing performance, especially when family firms are parts of
a business group. More remotely, our study is also related to the studies on managerial ownership
and firm performance (Morck, Shleifer, and Vishny, 1988). Again, in a business group setting, the
relationship between ownership and performance cannot be assessed without considering related-
party transactions.
Third, we add to the literature on business group tunneling (Bae, Kang, and Kim, 2002;
Bertrand, Mehta, and Mullainathan, 2002; Cheung, Rau, and Stouraitis, 2006; Baek, Kang, and
Lee, 2006). We report empirical evidence that related-party transactions between member firms
can be used as a tunneling vehicle benefiting founding family members at the expense of outside
-
- 6 -
minority shareholders. Our evidence on tunneling, however, is indirect in nature like in any other
tunneling papers.
This paper is organized as follows. Section 2 and 3 explain our research design, data, and key
variables. Section 4 reports our empirical results, and Section 5 concludes.
2. Research Design
In this paper, our aim is to quantify the effect of ownership change on firm’s related-party
transactions, earnings, dividend payouts, and its control over other firms. An obvious challenge to
this is the endogeneity of ownership change. To address this, difference-in-differences (DiD) or
instrumental variable (IV) approach making use of an exogenous shock to ownership change is in
order. But, unfortunately, we do not have such a shock in our sample – Korean chaebol firms
during 2000-2009.
So, we take a second-best approach of using covariate-matched control group firms. First, we
identify firm-years that experienced a major increase in heir’s ownership. We label this set of
firms as the treatment group. Second, for each treatment group firm, we identify its match among
firms that did not experience any major change in family ownership during our sample period and
that is from another chaebol group.4 We use three matching covariates: year, industry, and firm’s
asset size. Given the dominance of manufacturing firms in Korea, we use 4-digit Korea SIC code
for manufacturing and 2-digit for others. We label this set of firms as the control group, and
expect that the use of matching firms will significantly lower the risk of self-selection bias. Third,
by conducting difference-in-differences test, we compare these two groups of firms before and
after the treatment.
More specifically, we run the following regression to verify whether treatment group firms
experience greater increase in related-party transactions than control group firms after the
treatment.
4 Major family ownership changes include changes in heir’s net ownership by more than 5%p, changes in
controlling shareholder’s net ownership by more than 5%p, and changes in other relatives’ net ownership
by more than 5%p.
-
- 7 -
ittiitiitiit XTPTGTPTGRPT ενµβββα +++Φ+⋅+++= 210 (1)
itRPT is related-party transactions of firm i with other member firms at year t . We explain the
details of its measurement in the next section. iTG is a treatment group dummy that takes a value
of 1 if firm i is treated (experience a major increase in heir’s ownership during 2000-2009) and 0
otherwise. We explain what we exactly mean by ‘a major increase in heir’s ownership’ in the next
section. itTP is a treatment period dummy that takes a value of 1 if firm i is treated at year t or
before. Notice that this treatment period dummy is defined separately for each treatment group
firm i . Firm i and firm i ’s matching firm takes the same value for itTP . X is a column vector of
control variables. iµ and tν are respectively firm- and year-fixed effects. The coefficient of interest
is 2β , which we expect to be positive and statistically significant. Since same firms appear
multiple times in this panel regression, we use coefficient standard errors clustered at the firm-
level. Control variables include firm size, firm age, leverage, and a number of time-varying
dummies intended to capture abrupt changes in the volume of related-party transactions (spin-
offs, mergers, and new group affiliations). See Table 2, Panel A for their definitions.
To see whether the tendency of earnings increasing with related-party transactions strengthen
in treatment group firms after the treatment, we run the following regression with triple
interactions.
ittiititi
itititiitiititiit
XRPTTPTG
RPTTPRPTTGTPTGRPTTPTGEBITDA
ενµβ
ββββββα
+++Φ+⋅⋅+
⋅+⋅+⋅++++=
6
543210 (2)
itEBITDA is earnings before interest, tax, depreciation, and amortization of firm i at year t . We
explain the details of its measurement in the next section. Other variables are defined earlier. The
coefficient of interest is 6β , which we expect to be positive and statistically significant. Control
variables include firm size, firm age, leverage, cash holdings, R&D expenditure, and advertising
expenditure.
-
- 8 -
To see whether the tendency of dividend payout increasing with related-party transactions
strengthen in treatment group firms after the treatment, we run the following regression with
triple interactions.
ittiititi
itititiitiititiit
XRPTTPTG
RPTTPRPTTGTPTGRPTTPTGDIV
ενµβ
ββββββα
+++Φ+⋅⋅+
⋅+⋅+⋅++++=
6
543210 (3)
itDIV is cash dividend (including dividends paid out to preferred shareholders) paid out by firm i
at year t . We explain the details of its measurement in the next section. Other variables are
defined earlier. The coefficient of interest is 6β , which we expect to be positive and statistically
significant. Control variables include firm size, firm age, and leverage.
To see whether control over other member firm’s sensitivity to related-party transactions
strengthen in treatment group firms after the treatment, we run the following regression with
triple interactions.
ittiititi
itititiitiititiit
XRPTTPTG
RPTTPRPTTGTPTGRPTTPTGMCI
ενµβ
ββββββα
+++Φ+⋅⋅+
⋅+⋅+⋅++++=
6
543210 (4)
itMCI is marginal contribution to group control index of firm i at year t . We explain the details of
its measurement in the next section.
We conduct two falsification tests in this paper. First, we run difference-in-differences
regression (1) using treatment group firms where ‘non-heirs’ become a major shareholder. In this
regression, iTG takes a value of 1 if firm i experiences a major increase in non-heir’s ownership
during 2000-2009 and 0 otherwise. If the increase in related-party transactions are for the benefit
of heirs and their successions, they should not respond to changes in ‘non-heir’s’ ownership.
Second, we run difference-in-differences regressions (2) - (4), where counterparties of the
original treatment group firms (e.g. firms where heir becomes a major shareholder) are used as
our new treatment group. Again, if the increase in related-party transactions are for the benefit of
heirs and their successions, one should not see firms in the other side of transaction producing
-
- 9 -
higher earnings, paying out more dividends, or strengthening control over other firms.
3. Data and Key Variables
A. Sample Chaebol Groups
Our treatment and control group firms are from 26 chaebol groups that have been designated as
large business group by Korea Fair Trade Commission (KFTC) for at least 6 years during our
sample period of 2000-2009 (e.g. designated in the Aprils of 2001 to 2010). Table 1 lists the name
of 26 chaebol groups and the number of their member firms in each year. Since 1987, KFTC has
been designating large business groups and their member firms every year in April. Designation
depends on the aggregate size of member firms’ assets (net asset in case of financial firms),
measured at the end of prior year December. During 1993-2002, KFTC designated 30 largest
business groups without using any size threshold. During 2002-2008, KFTC used an explicit
threshold of 2 trillion Korean won and designated any group above the threshold as a large
business group. Since 2009, KFTC is using the threshold of 5 trillion won.
When announcing the list of large business groups, KFTC also announces the person who
controls the group and the list of firms under its control. For us, this is a very convenient feature
since we do not need to come up with an algorithm of our own to identify them. Control is
explicitly defined in the Monopoly Regulation and Fair Trade Act and its enforcement decree. It
considers not only shares directly owned, but also those indirectly owned through related parties,
such as relatives and other member firms. It also considers channels of influence that do not rely
on share ownership. A person in control can be both, a natural person or a legal person. In this
paper, we exclude the latter and focus on the former. For details on the identification of member
firms and the person in control, see Kim, Lim, and Sung (2007).
B. Major Increase in Heir’s Ownership
Since 2007, KFTC made public the detailed ownership structure of large business groups it
designates. This is done through a portal site, named OPNI, from which we download all the
necessary data for this paper. The data is available from 2000. When it comes to share ownership
-
- 10 -
among member firms, this data gives a complete picture. Complicated web of intragroup
ownership structure is summarized in a simple nn × matrix, where n is the number of member
firms. In this matrix, element ijx is the fraction of shares firm j owns in firm i .
But, when it comes to share ownership by controlling person’s family members, the data is
incomplete in a sense that it does not give information for each individual family member. For
privacy reasons, family owned shares are broken down into three groups: shares held by the
controlling shareholder (the person in control of the group), the immediate family members, and
the other relatives. Immediate family members include the spouse, the parents, and the children.
Other relatives include those within certain degrees of kinship (six with the controlling
shareholder or four with the spouse).
In this paper, we regard the shares held by immediate family members as those held by the
heir. There can be two potential problems for doing so. One is that it includes the shares held by
spouse and parents. Another has to do with the possibility of younger siblings, instead of children,
succeeding family ownership. The first problem is trivial since spouse and parents hardly own
shares.5 Among the treatment group firms we study in this paper, there is only one firm with
spouse’s ownership and none with parents’. In our robustness check, we obtain virtually the same
result after excluding this firm from the sample.6
The second problem is not a concern either since there are only a limited number of cases
where the group chairman position is succeeded by a younger sibling. A good example is Doosan,
where five brothers have taken turns in assuming the position. But, even in this case, shares have
not changed hands between brothers. Each brother inherited shares from their parents, and they
too are giving their shares to their respective children.
Our treatment group dummy iTG takes a value of 1 if firm i experiences a major increase in
heir’s ownership during 2000-2009 and 0 otherwise. A ‘major increase in heir’s ownership’
5 According to Economic Reform Research Institute (ERRI, 2012) the average (median) fraction of
spouse’s ownership out of that of immediate family is only 5.7% (0.1%) as of 2011 in case of top 20
chaebol groups. 6 Since 2009, each individual family member is required to disclose their detailed share ownership in each
member firm. In this paper, however, we do not make use of this data. At the time of this writing, this data
covers only four years, which is too short to investigate the key hypotheses of this paper.
-
- 11 -
means that its ‘net’ ownership (heir’s ownership – controlling shareholder’s ownership – other
relatives’ ownership) increases by more than 5%p and that its ownership is greater than those of
controlling shareholder’s and other relatives’. That is, HNOWN _Δ > p%5 , HOWN _ >
COWN _ , and HOWN _ > ROWN _ , where HNOWN _Δ is change in heir’s net ownership, and
HOWN _ , COWN _ , ROWN _ are respectively the ownership held by the heir, the controlling
shareholder, and other relatives.
Two points are worth noting here. First, we focus on ownership relative to other family
members (e.g. net ownership). If ownership increases not only for the heir, but also for others, the
subsequent increase in related-party transactions cannot be regarded as those just for the heir.
Likewise, the subsequent increase in firm’s importance in group control cannot be regarded as
that for the heir’s succession. By focusing on net ownership, we can effectively rule out such
alternative explanations. But, we do not exclude the possibility where the ownership of heir and
others both drop, but the drop of others is greater.7 Second, we impose a condition that the heir is
the largest shareholder among the family members. So, we exclude cases where heir’s net
ownership increases by more than 5%p, but its ownership is yet below that of other family groups.
Other treatment group dummies used in our falsification tests are similarly defined.
C. Marginal Contribution to Group Control
itMCI is the marginal contribution to group control index of firm i at year t . This index, originally
from Kim, Lim, and Sung (2007), is a measure devised to identify firms, through which a
controlling shareholder can most efficiently strengthen his control over other firms in the same
group. To be an efficient control vehicle, this firm must hold equity stakes in other firms, which
in turn hold equity stakes in others, which in turn hold equity stakes in others, and so on. One
way to quantify the degree of such direct and indirect equity holdings is to compute the cash flow
rights a controlling shareholder can additionally obtain from other firms when the vehicle firm
becomes a part of the group. Alternatively, one can compute the cash flow rights a controlling
shareholder will have to lose from others when the vehicle firm is no longer a group firm. By
7 Three such cases exist in our sample. If we drop them, statistical significance weakens, but our basic
results remain intact.
-
- 12 -
scaling this additional cash flow rights by the vehicle firm’s book equity, one can have a measure
that captures marginal contribution to group control. That is, the additional dollar amount of
equity one can obtain in other firms by investing one dollar of equity in the vehicle firm. Notice
that this measure does not identify firms that currently has the largest equity stakes in other firms.
Rather, it identifies firms that will become one in the future. The controlling shareholder that
wishes to maximize his control for a given amount of equity investment, has the incentive to
enlarge the firm with the highest MCI and let it grow into a firm that has the largest equity stakes
in other firms. Equation (4) shows the formula of firm i ’s marginal contribution to group control
index:
it
n
ijj
n
ijj
i
jtjtjtjt
itBE
cfrBEcfrBE
MCI
≠= ≠=
−−
=,1 ,1
(4)
itBE is firm i ’s book value of equity at year t . jtcfr is the cash flow rights controlling family has
in firm j when all member firms are included in the group. i
jtcfr−
is cash flow rights controlling
family has in firm j when all member firms, but firm i , are included in the group. The first term in
the numerator measures total cash flow rights the controlling family would receive from other
firms (denoted as j ) when firm i ( ji ≠ ) is included in the group. The second term in the
numerator captures total cash flow rights the control family would receive from other firms
(denoted as j ) when firm i ( ji ≠ ) is excluded from the group. We divide the difference by the
firm’s book equity to control for any size effect, since larger firms are more likely to have greater
contributions to group control.8
8 Our measure is similar, but not identical to the ‘centrality’ measure introduced by Almeida et al. (2011).
They identify firms by computing the average decrease in critical control threshold (CC) across all group
firms other than firm i, after excluding firm i from the group. Critical control (CC) threshold is the highest
control threshold that is consistent with family control of a firm. Control threshold T is the minimum votes
a family needs to hold directly or indirectly to control a firm. This measure has an advantage of using
voting rights, which we wish to capture, instead of cash flow rights. But, this measure is not adjusted for
firm size, and therefore has a tendency of favoring large firms that already has large control over others.
-
- 13 -
The index can have a value equal to zero. This happens when firm i does not have any equity
investment in other member firms. It should also be noted that the index has no upper bound. If
there is no restriction on debt or the length equity investment chain, the index can be well above
‘1.’ So, we winsorize the index at its 99th
percentile value.
Cash flow rights ( jtcfr ) is the sum of controlling family’s direct and indirect ownership.
Again, we follow Kim, Lim, and Sung (2007) and compute cash flow rights as follows:
⋅⋅⋅+++= = = =
n
k
n
k
n
l
lkljkkjkjjt dssdsdcfr1 1 1
(5)
jd is controlling family’s direct ownership in firm j at year t . Family includes the controlling
shareholder, its spouse, and relatives within certain degrees of kinship (six with the controlling
shareholder or four with the spouse). The subsequent terms are indirect ownership through
member firms under the control of the same controlling shareholder. For example, the second
term is family’s indirect ownership in firm j through firm k ( k can take values from 1 to n ). The
third term is family’s indirect ownership in firm j through firm k and firm l ( l can also take values
from 1 to n ). Since we know the intragroup ownership structure in a matrix form ( S ), a vector of
cash flow rights ( cfr ) can be easily computed by the following formula, where d is the vector of
direct family ownership.
( ) dSIcfr 1−−= (6)
D. Others
itRPT is the natural logarithm of 1 plus the sum of firm i ’s related-party sales to member firms at
year t plus firm i ’s related-party purchases from member firms at year t . Sales and purchases are
measured in million Korean won (approximately thousand US dollars) and adjusted for inflation
using Bank of Korea’s GDP deflator (base year = 2005). itRPS is the natural logarithm of 1 plus
the firm i ’s related-party sales to member firms at year t . itRPP is the natural logarithm of 1 plus
the firm i ’s related-party purchases from member firms at year t .
-
- 14 -
itEBITDA is the signed natural logarithm of earnings before interest, tax, depreciation, and
amortization of firm i at year t . A signed logarithm takes the logarithm of the absolute value of the
variable and assigns the original sign. Values for absolute value less than one are set to be zero.
Since we have many privately-traded firms in our sample, we use EBITDA instead of Tobin’s q
as our measure of firm performance. As is the case with related-party transactions, EBITDA is
also in million Korean won and adjusted for inflation using Bank of Korea’s GDP deflator (base
year = 2005). itDIV is the natural logarithm of 1 plus the cash dividend (including dividends paid
out to preferred shareholders) paid out to firm i at year t . Dividends are measured in million
Korean won and adjusted for inflation using Bank of Korea’s GDP deflator (base year = 2005).
Notice that we deliberately do not scale related-party transactions (RPT) or earnings by sales.
The focus of this paper is not on the fraction of RPT over sales nor on profit margins. Rather, we
are interested in the increase in earnings volume thanks to increase RPTs. Larger earnings matter
because it helps the firm to hold more equity stakes in other firms or payout more dividends to
the heir. But, we do control for size in our regressions. We include firm’s total assets in natural
logarithm as a covariate in our regressions. Total assets are also in million Korean won and
adjusted for inflation using Bank of Korea’s GDP deflator (base year = 2005).
The data on related-party transactions are available originally from each company’s business
reports (similar to 10K in US), but can be massively downloaded from KIS-Value, a financial
database administered by NICE Credit Information Service Co., Ltd. KIS-Value provides RPT
data not only for publicly-traded listed firms, but also for externally-audited private firms.
TS2000, a financial database administered by the Korea Listed Companies Association (KLCA),
provides RPT data limited to publicly-traded listed firms, but it gives the breakdown of RPT data
for each counterparty firm. itE and all other accounting variables used as controls are from
TS2000.
4. Results
A. A Preliminary Look
-
- 15 -
Table 2 gives the definition (Panel A) and the summary statistics (Panel B) of variables used in
this paper. Table 3 shows how changes in related-party transactions, earnings, and dividends vary
with changes in net ownership. It reports median (Pane A) and mean (Panel B) of paired-sample
differences in related-party transactions (RPTit, RPSit, RPPit), earnings (EBITDAit), and dividend
payouts (DIVit) before and after net ownership change of various degrees (±5%p, ±10%p, and
±15%p) for each family group (heir, controlling shareholder, and other relatives). The last three
columns report the average number of years before and after the net ownership change and the
number of firms used in the calculation for each level of ownership changes.
A number of observations can be made. First, the changes in related-party transactions and
earnings are positive, regardless of the level of net ownership change and the family member
being investigated. This is so even when they are all adjusted for inflation. Second, there is a
clear tendency of related-party transactions (RPTit, RPSit, RPPit) and earnings (EBITDAit)
increasing to a greater extent for larger changes in heir’s net ownership. Median changes in RPTit,
RPSit, RPPit, and EBITDAit respectively jump from 0.32, 0.16, 0.09, and 0.3 to 1.38, 1.13, 1.67,
and 1.11 as we move from the net ownership change of 15%p. The mean changes
look even more pronounced. Third, we can find a similar pattern for the controlling shareholder.
Median changes in RPTit, RPSit, RPPit, EBITDAit, and DIVit respectively jump from 0.59, 0.35,
0.59, 0.29, and 0.11 to 0.92, 0.58, 2.70, 0.55, and 0.36 as we move from the net ownership
change of 15%p. The jump, however, is much moderate than that for the heir with the
exception of related-party purchases. Similar pattern emerges when using ‘mean’ changes rather
than ‘median’ changes. Fourth, the patterns for other relatives are mixed. When using median
changes, there is a tendency of RPTit, RPSit, RPPit, EBITDAit, and DIVit increasing as we move
from the net ownership change of 15%p. But, such patter disappears when using mean
changes. Fifth, the changes in related-party transactions or earnings do not fall as we move from
the net ownership change of
-
- 16 -
Table 4 reports our first difference-in-differences (DiD) regression results. They are firm fixed
effects regressions of related-party transactions (RPTit, RPSit, and RPPit) on treatment group
dummy (TGi), treatment period dummy (TPit), their interaction (TGi x TPit), control variables,
and year dummies. We also include time-varying dummies capturing spin-offs, mergers, and new
affiliates, but their coefficients are suppressed. Treatment group dummy (TGi) is absorbed in firm
FE. Sample includes 36 treatment group firms that experienced major increase in heir’s
ownership and 36 control group firms identified by covariate matching based on year, industry
(4-digit code for manufacturing and 2-digit code for others), and asset size. t-values are reported
in the parenthesis, and are based on standard errors clustered at the firm level.
The coefficients of our interest (the coefficients on TGi x TPit) are all positive and marginally
significant. The coefficients of -0.2667 on TPit and 0.6021 on TGi x TPit in column (2) imply that
a firm experiencing a major increase in heir’s ownership also experiences a jump in related-party
sales by 0.3354 (=0.6021 – 0.2667), while its matching firm experiences a drop in related-party
sales by 0.2667. The difference-in-differences 0.6021 is approximately 6% of RPSit’s median
value (10.4). Among the controls, firm size is most significant.
C. Ownership Change, Related-Party Transactions, and Earnings
Table 5 reports our second difference-in-differences (DiD) regression results. They are firm fixed
effects regressions of earnings (EBITDAit) on treatment group dummy (TGi), treatment period
dummy (TPit), related-party transactions (RPTit, RPSit, or RPPit), their interactions, control
variables, and year dummies. Treatment group dummy (TGi) is absorbed in firm FE. Again,
sample includes 36 treatment group firms that experienced major increase in heir’s net ownership
and 36 control group firms identified by covariate matching based on year, industry (4-digit code
for manufacturing and 2-digit code for others), and asset size. t-values are reported in the
parenthesis, and are based on standard errors clustered at the firm level.
The coefficients of our interest (the coefficients on TGi x TPit x RPTit) are positive and
statistically significant when using related-party sales (column 2), positive and marginally
significant when using related-party transactions (sum of related-party sales and purchases), but
not significant when using related-party purchases (column 3). These results indicate that, it is
-
- 17 -
related-party sales, not related-party purchases, that benefit the treatment group firms after the
treatment. Purchasing raw materials and intermediate goods in favorable terms from other
member firms can be a source of profit. But, we do not see this effect in our sample. Rather, we
see firms benefiting greatly from sales of goods and services to other member firms.
This is consistent with the accusations made by non-governmental organizations (NGO) and
popular press against Korean chaebols.9 According to the reports they published, heirs benefit
from their equity stakes in firms that heavily rely on related-party sales to member firms. These
firms are mostly found in IT services, logistics, advertising, and constructions.
D. Ownership Change, Related-Party Transactions, and Dividend Payouts
Table 6 reports our third difference-in-differences (DiD) regression results. They are firm fixed
effects regressions of dividend payouts (DIVit) on treatment group dummy (TGi), treatment period
dummy (TPit), related-party transactions (RPTit, RPSit, or RPPit), their interactions, control
variables, and year dummies. Treatment group dummy (TGi) is absorbed in firm FE. Again,
sample includes 36 treatment group firms that experienced major increase in heir’s net ownership
and 36 control group firms identified by covariate matching based on year, industry (4-digit code
for manufacturing and 2-digit code for others), and asset size. t-values are reported in the
parenthesis, and are based on standard errors clustered at the firm level.
The coefficients of our interest (the coefficients on TGi x TPit x RPTit) are positive and
marginally significant when using related-party sales (column 2), but not significant when using
related-party transactions (column 1) or related-party purchases (column 3). These results
indicate that, it is related-party sales, not related-party purchases, that benefit the treatment group
firms after the treatment, and allow them to pay out more dividends. Again, this is consistent with
the accusations made by non-governmental organizations (NGO) and popular press that heirs
benefit from their equity stakes in firms that heavily rely on related-party sales to member firms.
E. Ownership Change, Related-Party Transactions, and Group Control
9 Solidarity for Economic Reform (SER) and its sister organization, Economic Reform Research Institute
(ERRI), are the two pioneering NGOs in this area. Since 2006, they have been publishing a number of
reports on related-party sales aimed to benefit controlling family members.
-
- 18 -
Table 7 reports our fourth difference-in-differences (DiD) regression results. They are firm fixed
effects regressions of marginal contribution to group control index (MCIit) on treatment group
dummy (TGi), treatment period dummy (TPit), related-party transactions (RPTit, RPSit, or RPPit),
their interactions, and year dummies. Treatment group dummy (TGi) is absorbed in firm FE.
Again, sample includes 36 treatment group firms that experienced major increase in heir’s net
ownership and 36 control group firms identified by covariate matching based on year, industry
(4-digit code for manufacturing and 2-digit code for others), and asset size. t-values are reported
in the parenthesis, and are based on standard errors clustered at the firm level.
The results on marginal contribution to group control are very similar to those on earnings.
The coefficients of our interest (the coefficients on TGi x TPit x RPTit) are positive and
statistically significant when using related-party sales (column 2), positive and marginally
significant when using related-party transactions (sum of related-party sales and purchases), but
not significant when using related-party purchases (column 3). Again, it is related-party sales, not
related-party purchases, that benefit treatment group firms in terms of control over other firms.
This is also consistent with the anecdotal evidence reported by NGOs and popular press.
F. Falsification Tests
Table 8 reports the results of our first falsification test. We run difference-in-differences
regressions identical to those reported in Table 4, but using treatment group firms where ‘non-
heirs’ become a major shareholder. Columns (1) – (3) ((4) – (6)) use 25 (30) treatment group
firms that experienced major increase in controlling shareholder’s (other relative’s) ownership
and 25 (30) control group firms identified by covariate matching based on year, industry (4-digit
code for manufacturing and 2-digit code for others), and asset size. If increase in related-party
transactions are for the benefit of heirs and their successions, it should not respond to changes in
‘non-heir’s’ ownership. This is what we find. The coefficients of our interest (the coefficients on
TGi x TPit) are all positive, but statistically insignificant, indicating that related-party transactions
do not increase in firms where ‘non-heirs’ become a major shareholder.
In our second falsification test, we run difference-in-differences regressions identical to those
reported in Tables 5-7, but using counterparties of the original treatment group firms (e.g. firms
-
- 19 -
where heir become a major shareholder) as our new treatment group. Again, if increase in related-
party transactions are for the benefit of heirs and their successions, one should not see firms in
the other side of transaction producing higher earnings or strengthening control over other firms.
For this second falsification test, we make use of TS2000, the financial database administered
by Korea Listed Companies Association (KLCA), which provides the name, but not the code of
counterparties (including privately-traded firms) that engage in related-party transactions with
listed firms. Among these counterparties, we exclude individuals and overseas subsidiaries. The
remaining firms become our new treatment group firms. For each treatment group firm, we again
identify matching firms based on year, industry, and asset size. But, we do this only during 2000-
2004, where we have identified the code of counterparties. We plan to extend the sample period
so that it covers that of our key regressions. Presently, we have only 18 firms in our treatment
group.
Table 9 reports the results. iTG takes a value of 1 if firm i is the counterparty of the original
treatment group firm, and 0 otherwise. itTP takes a value of 1 if the original treatment group firm
of firm i experiences a major increase in heir’s net ownership at year t or before. All related-party
transaction variables are defined from the perspective of the counterparty firm. The coefficients
of our interest (the coefficients on TGi x TPit x RPTit) are all statistically insignificant, indicating
that related-party transactions do not improve the counterparty firms’ earnings, dividend payouts,
or strengthen their control over other firms.
5. Conclusion
In this paper, we investigate if families controlling business groups make use of related-party
transactions to benefit firms, in which heirs hold significant equity stakes, and thereby let such
firms grow large enough to strengthen their control over other firms within the group. Using a
sample of Korean chaebol firms during 2000-2009, we report a number of results consistent with
our hypotheses. First, related-party transactions increase in firms where heirs become a major
shareholder (treatment group) after the ownership change (treatment). Second, earnings increase
with related-party transactions in treatment group firms after the treatment. Third, dividend
-
- 20 -
payout increase with related-party transactions in treatment group firms after the treatment.
Fourth, importance in group control increases with related-party transactions in treatment group
firms after the treatment.
These academic findings confirm the non-academic allegations made by non-governmental
organizations (NGOs) and popular press in Korea. It also justifies the new regulatory actions
taken by the Korean government in recent years to curb tunneling. In December 2011, the
National Assembly passed a bill revising the Inheritance and Gift Tax Act and allowing the
National Tax Office to levy gift tax on expropriated income from related party sales. More
specifically, shareholders individually owning more than 3% (directly or indirectly) of total
outstanding shares of a company, where related-party sales take up more than 30% of its total
sales, are subject to a gift tax. The taxable gift income is equal to earnings before tax (NOPLAT)
× (percentage of related-party sales out of total sales – 15%) × (percentage of shareholding – 3%).
Another regulatory action took place in August 2013. The National Assembly passed a bill
revising the Monopoly Regulation and Fair Trade Act and allowing the Fair Trade Commission
to levy penalty on related-party transactions favoring controlling family members. The new rule
applies to members of large business groups designated by KFTC. To be identified as a
beneficiary firm, controlling family members in aggregate must directly own more than 30
percent of outstanding shares and must have entered related-party transactions in significantly
favorable terms.
We believe our findings are relevant not only to Korea, but also to many other countries.
Family controlled business groups are prevalent in emerging markets and even in some
developed economies (Khanna and Yafeh, 2007). Families controlling these business groups may
use related-party transactions as means of family ownership succession if the country’s
regulatory environment does not permit an easy solution to it.
This paper contributes to the literature in three main ways. First, to the best of our knowledge,
this is the first paper on ‘family ownership’ succession, which we believe is an issue of
paramount importance for family firms, but its research virtually missing in the existing literature.
As mentioned earlier, the main focus of existing papers is on ‘managerial’ succession. Second, we
contribute to the studies on family firm performance, which became popular during the past
several years. We contribute to this area of study by highlighting the importance of related-party
-
- 21 -
transactions when assessing firm performance, especially when firms are parts of a business
group. Third, we add to the literature on tunneling among business group firms. We report
empirical evidence that related-party transactions can be used as a tunneling vehicle benefiting
founding family members at the expense of outside minority shareholders. Our evidence on
tunneling, however, is indirect in nature like in any other tunneling papers.
-
- 22 -
References
Almeida, Heitor, Sang Yong Park, Marti Subrahmanyam, and Daniel Wolfenzon (2011), “The Structure and Formation of Business Groups: Evidence from Korean Chaebols,” Journal of Financial Economics, Vol.99, No.2, pp.447-475
Anderson, Ronald and David Reeb (2003), “Founding-Family Ownership and Firm Performance: Evidence from the S&P 500,” The Journal of Finance, Vol.58, pp.1301-1327
Andres, Christian (2008), “Large Shareholders and Firm Performance – An Empirical Examination of
Founding-Family Ownership,” Journal of Corporate Finance, Vol.14, pp.431-445
Bae, Kee-Hong, Jun-Koo Kang, and Jin-Mo Kim (2002), “Tunneling or Value Added? Evidence from Mergers by Korean Business Groups,” The Journal of Finance, Vol.57 No.6, pp.2695-2740
Baek, Jae-Seung, Jun-Koo Kang, and Inmoo Lee (2006), “Business Groups and Tunneling: Evidence from Private Securities Offerings by Korean Chaebols,” The Journal of Finance, Vol.61 No.5, pp.2415-2449
Bennedsen, Morten, Kasper Nielsen, Francisco Perez-Gonzales, and Daniel Wolfenzon (2007), “Inside the Family Firm: The Role of Families in Succession Decisions and Performance,” Quarterly Journal of Economics, Vol.122 No.2, pp.647-691
Bertrand, Marianne, Paras Mehta, and Sendhil Mullainathan (2002), “Ferreting Out Tunneling: An Application to Indian Business Groups,” Quarterly Journal of Economics, Vol.117, pp.121-148
Cheung, Yan-Leung, Raghavendra Rau, and Aris Stouraitis (2006), “Tunneling, Propping, and Expropriation: Evidence from Connected Party Transactions in Hong Kong,” Journal of Financial Economics, Vol.82 No.2, pp.343-386
Claessens, Stijn, Simeon Djankov, and Larry Lang (2000), “The Separation of Ownership and Control in East Asian Corporations,” Journal of Financial Economics, Vol.58, pp.81-112
Cucculelli, Marco and Giancinto Micucci (2008), “Family Succession and Firm Performance: Evidence
from Italian Family Firms,” Journal of Corporate Finance, Vol.14, pp.17-31
Economic Reform Research Institute (2012), “Family Ownership Succession of Chaebol Groups,”
Economic Reform Report No. 2012-17 (in Korean)
Faccio, Mara and Larry H.P. Lang (2002), “The Ultimate Ownership of Western European Corporations,” Journal of Financial Economics Vol.65, pp. 365-395
Gompers, Paul., Joy Ishii, and Andrew Metrick (2010), “Extreme Governance: An Analysis of Dual-Class Firms in the United States, Review of Financial Studies, Vol.23 No.3, pp.1051-1088.
Khanna, Tarun and Yishay Yafeh (2007), “Business Groups in Emerging Markets: Paragons or Parasites?”
Journal of Economic Literature, Vol.45 No2., pp.331-371
-
- 23 -
Kim, Woochan, Youngjae Lim, and Taeyoon Sung (2007), “Group Control Motive as a Determinant of Ownership Structure in Business Conglomerates: Evidence from Korea’s Chaebols,” Pacific-Basin Finance Journal, Vol.15 No.3, pp.213-252
Lopez de Silanes, Rafael La Porta, and Andrei Shleifer (1999), “Corporate Ownership around the World,”
The Journal of Finance Vol.54 No.2 pp.471-517
Maury, Benjamin (2006), “Family Ownership and Firm Performance: Empirical Evidence from Western
European Corporations,” Journal of Corporate Finance, Vol.12, pp.321-341
McConaughy, Daniel, Michael Walker, Glenn Henderson, and Chandra Mishra (1998), “Founding Family
Controlled Firms: Efficiency and Value, Review of Financial Economics, Vol.7, No.1, pp.1-19
Miller, Danny, Isabelle Le Breton-Miller, Richard Lester, and Albert Cannella (2007), “Are Family Firms
Really Superior Performers?” Journal of Corporate Finance, Vol.13, pp.829-858
Morck, Randall, Andrei Shleifer, and Robert Vishny (1988), “Management Ownership and Market Valuation: An Empirical Analysis,” Journal of Financial Economics, Vol.20, pp.293-315
Nenova, Tatiana (2003), “The Value of Corporate Voting Rights and Control: A Cross-Country Analysis,” Journal of Financial Economics, Vol.68, pp.325-351
Smith, Brian and Ben Amoako-Adu (1999), “Management Succession and Financial Performance of
Family Controlled Firms,” Journal of Corporate Finance, Vol.5, pp.341-368
Villalonga, Belen and Raphael Amit (2006), “How Do Family Ownership, Control, and Management Affect Firm Value?” Journal of Financial Economics, Vol.80, No.2, pp.385-417
Villalonga, Belen and Rephael Amit (2009), “How are U.S. Family Firms Controlled?” Review of
Financial Studies, Vol. 22, No.8, pp.3047-3091
-
Fig
Following
related-pa
gure 1: Hanw
g figure depic
arty sales volu
wha S&C’s U
cts how major
ume, earnings,
Use of Relat
r change in H
and control o
- 24 -
ted-Party Tr
Hanwha S&C
ver other mem
ransactions f
’s ownership
mber firms in t
for Tunnelin
structure pro
the group.
ng and Succe
mpted the inc
ession
crease in its
-
25
Table 1: Sample Chaebol Groups
List of chaebol groups, and the number of their member firms in each year. Column (1) lists a total of 26 Chaebols designated by the FTC for at least 6 years during the
sample period of 2000-2009 (e.g. designated in the Aprils of 2001 to 2010). Column (2) counts the number of member firms in each group in each year. Column (3) shows
the controlling shareholders’ names and column (4) the generations from founders (1, 2, and 3 indicates 1st , 2nd, and 3rd generation).
No
(1) (2) (3) (4)
Chaebol Name Number of Member Firms
Controlling Shareholder Gener-
ation 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
1 CJ 26 24 30 38 45 53 60 62 61 53 Lee Jae-Hyun 3
2 Daelim 14 13 13 11 11 12 14 14 16 16 Lee Joon-Yong 2
3 Dongbu 13 15 17 16 14 15 15 21 24 22 Kim Jun-Ki 1
4 Dongkuk Steel 8 6 7 8 8 12 11 12 13 11 Jang Se-Joo 3
5 Doosan 16 16 20 20 17 17 17 19 22 25 Park Yong-Gon 3
6 GS - - - - 40 49 48 57 63 68 Huh Chang-Soo 3
7 Hanjin 17 19 21 21 22 21 25 27 32 37 Cho Joong-Hoon (~2002), Cho Yang-Ho (2003~) 1, 2
8 Hanwha 21 22 27 25 25 24 27 33 35 40 Kim Seung-Youn 2
9 Hyosung 14 14 14 15 15 16 22 28 39 38 Cho Suk-Rae 2
10 Hyundai 18 11 9 8 6 9 8 8 9 11 Chung Mong-Hun (~2003), Hyun Jeong-Eun (2004~) 2, 3
11 Hyundai Department Store 15 10 19 17 20 23 24 25 24 29 Chung Mong-Keun (~2006), Chung Ji-Sun (2007~) 2, 3
12 Hyundai Development Company 8 9 10 11 11 12 15 14 15 14 Chung Se-Young (~2006), Chung Mong-Kyu (2007~) 1, 2
13 Hyundai Heavy Industries - - 3 3 4 4 4 6 8 10 Chung Mong-Joon 2
14 Hyundai Motor Company 14 21 21 26 26 37 34 33 37 38 Chung Mong-Koo 2
15 KCC - 5 5 6 4 4 4 6 9 9 Chung Sang-Yong 1
16 Kolon 22 28 31 30 27 22 33 33 37 35 Lee Dong-Chan (~2006), Lee Woong-Yeol (2007~) 2, 3
17 Kumho 14 13 13 14 16 21 34 50 47 46 Park Sung-Yong (~2005), Park Sam-Koo (2006~) 2, 2
18 LG 37 46 45 45 47 32 31 36 52 53 Koo Bon-Moo 3
19 Lotte 30 31 33 33 39 41 42 43 51 57 Shin Kyuk-Ho 1
20 LS - - - 7 17 19 19 23 31 43 Koo Tae-Hoi 1
21 OCI 22 19 19 19 18 19 18 15 18 18 Lee Hoi-Rim (~2007), Lee Soo-Young (2008~) 1, 2
22 Samsung 55 54 54 55 53 49 50 49 53 57 Lee Kun-Hee 2
23 Shinsegae 9 10 12 12 13 14 15 15 15 12 Lee Myung-Hee 2
24 SK 50 57 55 54 46 54 55 63 75 74 Chey Tae-Won 2
25 Tongyang 20 8 7 8 8 8 14 13 15 17 Hyun Jae-Hyun 2
26 Youngpoong 23 24 23 20 19 26 22 21 22 23 Jang Byung-Hee (~2002), Jang Hyung-Jin (2003~) 1, 2
Total number of Chaebols 22 23 24 25 26 26 26 26 26 26
Total number of group firms 466 475 508 522 571 613 661 726 823 856
-
26
Table 2: Variable Definitions and Summary Statistics
Definition and summary statistics of variables used in this paper. Panel A defines each variable. Panel B
provides summary statistics for each family group. We use nonfinancial firms from 26 chaebol groups (see Table
1) during 2000-2009.
Panel A. Variable Definitions
Variables Definitions
Left-Hand Side Variables
RPTit ln[(sum of related-party sales and purchases/ GDP deflator) + 1]; related-party
sales and purchase are measured in million Korean won (approximately
thousand US dollars)
RPSit ln[(related-party sales/ GDP deflator) + 1]; related-party sales are measured in
million Korean won (approximately thousand US dollars)
RPPit ln[(related-party purchases/ GDP deflator) + 1]; related-party purchase are
measured in million Korean won (approximately thousand US dollars)
EBITDAit ln(absolute value of earnings before interest, taxes, depreciation and
amortization/GDP deflator) x sign of original EBITDA if its absolute value is
greater than 1, and 0 otherwise.
MCIit Marginal contribution to group control index; winsorized at the 99th percentile
values; see Section 3.C for the details of its definition.
DIVit ln[(dividend/ GDP deflator) + 1]; dividend is measured in million Korean won
(approximately thousand US dollars)
Right-Hand Side Variables
TGi Treatment group dummy, which takes a value of 1 if the firm i experiences a
major increase in heir’s ownership and 0 otherwise. Major increase in heir’s
ownership means that its net ownership (heir’s ownership – controlling
shareholder’s ownership – other relatives’ ownership) increases by more than
5%p and it is greater than both, the controlling shareholder’s ownership and
the other relatives’ ownership. Other treatment group dummies used in our
falsification tests are similarly defined.
TPit Treatment period dummy that takes a value of 1 if firm i is treated at year t or
before. Notice that this treatment period dummy is defined separately for each
treatment group firm i. Firm i and firm i’s matching firm takes the same value
for TPit.
NOWN_Hit Heir’s net ownership; heir’s ownership (OWN_Hit) – controlling shareholders’
ownership (OWN_Cit) – other relatives’ ownership (OWN_Rit)
NOWN_Cit Controlling shareholders’ net ownership; controlling shareholders’ ownership
(OWN_Cit) – heir’s ownership (OWN_Hit) – other relatives’ ownership
(OWN_Rit)
NOWN_Rit Other relatives’ net ownership; other relatives’ ownership (OWN_Rit) – heir’s
ownership (OWN_Hit) – controlling shareholders’ ownership (OWN_Cit)
Firm size ln(Total assets/GDP deflator); total assets are measured in million Korean
won (approximately thousand US dollars)
Firm age Number of years since a firm’s establishment, measured by ln(year - year of
establishment)
Leverage ln[(Book value of debt /total assets)+1]
Cash holdings ln(Cash and cash equivalents / total assets)
R&D expenditure ln[(R&D/ Sales) x 100 +1]; winsorized at the 99th percentile values
Advertising expenditure ln[(Advertising / Sales) x 100 +1]; winsorized at the 99th percentile values
Spin-off 1 if a firm experiences spin-off at year t or before, and 0 otherwise.
Merger 1 if a firm experiences merger at year t or before, and 0 otherwise.
New affiliate 1 if a firm is affiliated to a business group at year t or before, and 0 otherwise.
-
27
Panel B. Summary Statistics
Variables Heir Controlling Shareholders Other Relatives
N Mean Med. S.D. Min. Max. N Mean Med. S.D. Min. Max. N Mean Med. S.D. Min. Max.
RPTit 632 10.85 10.95 1.95 0.00 15.47 416 10.66 10.75 2.19 0.00 14.67 507 9.64 9.74 2.86 0.00 14.93
RPSit 632 10.02 10.36 2.56 0.00 15.18 416 9.32 10.14 3.12 0.00 14.29 507 8.49 9.29 3.45 0.00 14.06
RPPit 632 8.98 9.32 2.72 0.00 14.70 416 9.09 9.54 3.13 0.00 14.03 507 7.87 8.19 3.64 0.00 14.68
EBITDAit 617 7.26 9.00 5.85 -11.49 14.42 384 8.17 9.25 5.31 -13.16 13.94 495 6.27 8.37 6.34 -13.16 13.64
MCIit 626 0.26 0.00 0.72 0.00 3.98 418 0.23 0.00 0.65 0.00 3.98 459 0.28 0.00 0.59 0.00 3.98
DIVit 616 3.93 0.00 4.17 0.00 12.38 384 4.14 0.00 4.38 0.00 11.58 495 2.75 0.00 3.89 0.00 11.58
TGi 650 0.51 1.00 0.50 0.00 1.00 430 0.53 1.00 0.50 0.00 1.00 538 0.50 0.50 0.50 0.00 1.00
TPit 650 0.49 0.00 0.50 0.00 1.00 430 0.50 1.00 0.50 0.00 1.00 538 0.54 1.00 0.50 0.00 1.00
NOWNit 637 0.02 0.00 0.19 -0.60 1.00 430 -0.03 0.00 0.09 -0.44 0.20 482 -0.11 0.00 0.33 -1.00 0.75
OWNit 637 0.07 0.00 0.15 -0.03 1.00 430 0.03 0.00 0.07 -0.01 0.43 482 0.13 0.00 0.29 -0.41 1.00
Firm size 617 11.99 12.01 1.52 8.66 15.86 384 12.32 12.18 2.10 6.45 17.03 495 11.68 11.16 1.88 8.26 15.95
Firm age 629 2.45 2.64 0.96 0.00 4.09 415 2.48 2.64 1.08 0.00 4.03 516 2.32 2.30 1.08 0.00 4.33
Leverage 617 0.43 0.44 0.13 0.01 0.72 384 0.43 0.46 0.15 0.01 0.76 495 0.47 0.49 0.17 0.00 1.16
Cash holdings 617 0.06 0.04 0.07 0.00 0.51 384 0.06 0.03 0.09 0.00 0.64 495 0.06 0.02 0.09 0.00 0.51
R&D 650 0.08 0.00 0.27 0.00 1.42 430 0.12 0.00 0.32 0.00 1.42 538 0.05 0.00 0.20 0.00 1.42
Advertising 650 0.30 0.04 0.57 0.00 2.88 430 0.23 0.04 0.39 0.00 2.56 538 0.35 0.06 0.51 0.00 2.88
Spin-off 650 0.00 0.00 0.00 0.00 0.00 430 0.00 0.00 0.00 0.00 0.00 538 0.00 0.00 0.04 0.00 1.00
Merger 650 0.00 0.00 0.00 0.00 0.00 430 0.01 0.00 0.12 0.00 1.00 538 0.00 0.00 0.00 0.00 0.00
New Affiliate 650 0.00 0.00 0.00 0.00 0.00 430 0.00 0.00 0.00 0.00 0.00 538 0.02 0.00 0.14 0.00 1.00
-
28
Table 3: Changes in RPT and EBITDA Before and After Ownership Changes
The median (Panel A) and the mean (Panel B) of paired-sample differences in RPTit, RPSit, RPPit, and EBITDAit
before and after net ownership change of various degrees (±5%p, ±10%p, and ±15%p) for each family group
(heir, controlling shareholder, and other relatives). See Table 2 for the definitions of RPTit, RPSit, RPPit, and
EBITDAit. The year of net ownership change is not used in the computation. Sample consists of nonfinancial
firms from 26 chaebol groups (see Table 1) during 2000-2009. We exclude firms that underwent a spin-off or a
merger from the sample. We also exclude newly added member firms during the sample period.
Panel A: Median of paired-sample differences
△Ownership Before △Ownership - After △Ownership no. of years
(before)
no. of
years
(after)
no. of
firmsRPTit RPSit RPPit EBITDAit DIVit
NOWN_Hit
> 15%p 1.38 1.13 1.67 1.11 0.14 3.7 3.8 22
> 10%p 1.11 0.88 1.60 0.98 0.00 3.6 3.7 27
> 5%p 0.64 0.87 0.80 0.63 0.00 3.5 3.8 35
< -5%p 0.32 0.16 0.09 0.30 0.72 3.6 3.3 37
< -10%p 0.41 0.10 0.21 0.38 0.12 3.4 3.7 24
< -15%p 0.64 0.20 0.51 0.46 0.11 3.0 3.1 16
NOWN_Cit
> 15%p 0.92 0.58 2.70 0.55 0.36 5.5 3.0 4
> 10%p 0.41 0.28 0.22 0.30 0.49 4.6 3.1 10
> 5%p 0.48 0.66 0.23 0.31 0.99 4.6 3.1 17
< -5%p 0.59 0.35 0.59 0.29 0.11 3.9 3.4 49
< -10%p 0.90 0.39 0.76 0.36 0.14 4.1 3.0 35
< -15%p 1.24 0.40 0.91 0.63 0.27 4.2 3.1 26
NOWN_Rit
> 15%p 0.66 0.39 0.44 0.62 0.31 3.8 2.7 13
> 10%p 0.66 0.39 0.44 0.62 0.31 3.9 3.2 13
> 5%p 0.26 0.16 0.65 0.19 1.35 3.8 3.1 18
< -5%p 0.61 0.68 0.49 0.47 0.09 3.5 4.0 42
< -10%p 0.77 0.32 0.56 0.58 0.13 3.2 4.4 28
< -15%p 1.22 0.95 0.80 0.84 0.05 3.3 4.1 21
Panel B: Mean of paired-sample differences
△Ownership Before △Ownership - After △Ownership no. of years
(before)
no. of
years
(after)
no. of
firmsRPTit RPSit RPPit EBITDAit DIVit
NOWN_Hit
> 15%p 2.73 2.33 3.25 3.85 1.34 3.7 3.8 22
> 10%p 2.17 1.85 2.99 3.28 0.83 3.6 3.7 27
> 5%p 1.64 1.57 2.20 2.41 0.35 3.5 3.8 35
< -5%p 0.84 0.29 0.55 0.85 1.23 3.6 3.3 37
< -10%p 1.05 0.25 0.64 1.15 0.70 3.4 3.7 24
< -15%p 1.64 0.36 1.36 1.37 0.49 3.0 3.1 16
NOWN_Cit
> 15%p 0.85 0.90 1.53 2.09 0.88 5.5 3.0 4
> 10%p 0.60 0.66 0.95 0.12 1.33 4.6 3.1 10
> 5%p 0.68 1.06 0.88 1.88 1.83 4.6 3.1 17
< -5%p 1.13 0.43 1.09 1.24 0.09 3.9 3.4 49
< -10%p 1.56 0.56 1.53 1.73 -0.01 4.1 3.0 35
< -15%p 1.99 0.56 1.84 2.56 0.87 4.2 3.1 26
NOWN_Rit
> 15%p 1.83 1.63 1.39 1.92 2.12 3.8 2.7 13
> 10%p 1.65 1.49 1.19 1.82 1.99 3.9 3.2 13
> 5%p 1.27 1.06 1.34 1.57 2.60 3.8 3.1 18
< -5%p 0.84 0.66 0.96 1.40 1.07 3.5 4.0 42
< -10%p 0.79 0.39 0.81 1.32 1.54 3.2 4.4 28
< -15%p 1.02 0.56 1.01 2.69 1.24 3.3 4.1 21
-
29
Table 4: Ownership Change and Related-Party Transactions
Firm fixed effects regressions of related-party transactions (RPTit, RPSit, and RPPit) on treatment group dummy
(TGi), treatment period dummy (TPit), their interaction (TGi x TPit), control variables, and year dummies. We
also include time-varying dummies capturing spin-offs, mergers, and new affiliates, but their coefficients are
suppressed. Treatment group dummy (TGi) is absorbed in firm FE. Sample include 36 treatment group firms that
experienced major increase in heir’s ownership and 36 control group firms identified by covariate matching
based on year, industry (4-digit code for manufacturing and 2-digit code for others), and asset size. t-values are
reported in the parenthesis, and are based on standard errors clustered at the firm level. *, **, and ***
respectively indicate significance at 10%, 5%, and 1% levels. Significant results (at 10% level or better) are
shown in boldface.
(1) (2) (3)
Related-party
Transactions (RPTit)
Related-party
Sales (RPSit)
Related-party
Purchases (RPPit)
TPit -0.1324 -0.2667 -0.2387
(-0.93) (-1.38) (-0.92)
TGi x TPit 0.3906* 0.6021* 0.6158*
(1.80) (1.67) (1.75)
Firm size 0.6822*** 0.3782 1.0067*** (3.23) (1.50) (3.48)
Firm age 0.3279* 0.6744* 0.2891
(1.72) (1.84) (0.76)
Leverage -0.3840 0.2795 -0.6671
(-0.34) (0.23) (-0.39)
Constant Yes Yes Yes
Firm FE Yes Yes Yes
Year FE Yes Yes Yes
Observations 602 602 602
Number of firms 72 72 72
within R-sq 0.192 0.136 0.167
-
30
Table 5: Ownership Change, Related-Party Transactions, and Earnings
Firm fixed effects regressions of earnings (EBITDAit) on treatment group dummy (TGi), treatment period
dummy (TPit), related-party transactions (RPTit, RPSit, or RPPit), their interactions, control variables, and year
dummies. Treatment group dummy (TGi) is absorbed in firm FE. Sample include 36 treatment group firms that
experienced major increase in heir’s ownership and 36 control group firms identified by covariate matching
based on year, industry (4-digit code for manufacturing and 2-digit code for others), and asset size. t-values are
reported in the parenthesis, and are based on standard errors clustered at the firm level. *, **, and ***
respectively indicate significance at 10%, 5%, and 1% levels. Significant results (at 10% level or better) are
shown in boldface.
(1) (2) (3)
Dependent Var. EBITDAit
Types of RPT Related-Party
Transactions (RPTit)
Related-party
Sales (RPSit)
Related-party
Purchases (RPPit)
TPit 6.6177* 6.1879*** 4.3488*
(1.84) (2.79) (1.75)
RPTit 0.7819 0.6752*** 0.5615*
(1.64) (3.31) (1.99)
TGi x TPit -9.8245* -8.4469** -4.1508
(-1.89) (-2.51) (-1.52)
TGi x RPTit -0.2157 -0.3760 -0.2650
(-0.34) (-1.14) (-0.76)
TPit x RPTit -0.5789* -0.5740** -0.4506*
(-1.78) (-2.60) (-1.85)
TGi x TPit x RPTit 0.7650* 0.6947** 0.3094
(1.71) (2.15) (1.11)
Firm size 0.5215 0.9003 0.4839
(0.70) (1.29) (0.64)
Firm age 2.6736*** 2.5072*** 2.6968*** (3.12) (3.07) (2.95)
Leverage -5.3704 -6.0444 -5.3130
(-1.38) (-1.64) (-1.42)
Cash holdings 2.8052 2.7519 4.6822
(0.60) (0.61) (0.94)
R&D expenditure -0.4417 -0.5094 -0.3162
(-0.39) (-0.49) (-0.30)
Advertising expenditure -2.3931** -2.5513** -2.2672**
(-2.12) (-2.12) (-2.33)
Constant Yes Yes Yes
Firm FE Yes Yes Yes
Year FE Yes Yes Yes
Observations 602 602 602
Number of firms 72 72 72
within R-sq 0.102 0.112 0.0956
-
31
Table 6: Ownership Change, Related-Party Transactions, and Dividends
Firm fixed effects regressions of dividends (DIVit) on treatment group dummy (TGi), treatment period dummy
(TPit), related-party transactions (RPTit, RPSit, or RPPit), their interactions, control variables, and year dummies.
Treatment group dummy (TGi) is absorbed in firm FE. Sample include 36 treatment group firms that
experienced major increase in heir’s ownership and 36 control group firms identified by covariate matching
based on year, industry (4-digit code for manufacturing and 2-digit code for others), and asset size. t-values are
reported in the parenthesis, and are based on standard errors clustered at the firm level. *, **, and ***
respectively indicate significance at 10%, 5%, and 1% levels. Significant results (at 10% level or better) are
shown in boldface.
(1) (2) (3)
Dependent Var. DIVit
Types of RPT Related-Party
Transactions (RPTit)
Related-party
Sales (RPSit)
Related-party
Purchases (RPPit)
TPit 0.4972 1.5026 0.0104
(0.33) (1.37) (0.01)
RPTit 0.1426 0.1693* 0.0893
(0.74) (1.84) (0.74)
TGi x TPit -2.2009 -3.7044*** -0.7998
(-1.14) (-2.72) (-0.48)
TGi x RPTit 0.1695 -0.0175 0.0007
(0.63) (-0.11) (0.00)
TPit x RPTit -0.0276 -0.1271 0.0144
(-0.21) (-1.33) (0.09)
TGi x TPit x RPTit 0.0811 0.2402* -0.0413
(0.47) (1.87) (-0.24)
Firm size 0.3718 0.4890 0.4414
(1.17) (1.52) (1.34)
Firm age 1.0798** 1.0415** 1.0873** (2.20) (2.20) (2.18)
Leverage -3.5381** -3.6166** -3.4048** (-2.11) (-2.23) (-2.03)
Constant Yes Yes Yes
Firm FE Yes Yes Yes
Year FE Yes Yes Yes
Observations 602 602 602
Number of firms 72 72 72
within R-sq 0.124 0.127 0.117
-
32
Table 7: Ownership Change, Related-Party Transactions, and Group Control
Firm fixed effects regressions of marginal contribution to group control index (MCIit) on treatment group
dummy (TGi), treatment period dummy (TPit), related-party transactions (RPTit, RPSit, or RPPit), their
interactions, and year dummies. Treatment group dummy (TGi) is absorbed in firm FE. Sample include 36
treatment group firms that experienced major increase in heir’s ownership and 36 control group firms identified
by covariate matching based on year, industry (4-digit code for manufacturing and 2-digit code for others), and
asset size. t-values are reported in the parenthesis, and are based on standard errors clustered at the firm level. *,
**, and *** respectively indicate significance at 10%, 5%, and 1% levels. Significant results (at 10% level or
better) are shown in boldface.
(1) (2) (3)
Dependent Var. MCIit
Types of RPT Related-Party
Transactions (RPTit)
Related-party
Sales (RPSit)
Related-party
Purchases (RPPit)
TPit 0.1457 0.1285 0.0577
(1.10) (1.29) (0.49)
RPTit 0.0013 0.0006 0.0019
(0.19) (0.15) (0.36)
TGi x TPit -0.4203 -0.4640* -0.0036
(-1.24) (-1.93) (-0.02)
TGi x RPTit -0.0132 -0.0259 0.0140
(-0.58) (-1.47) (0.60)
TPit x RPTit -0.0127 -0.0121 -0.0049
(-1.06) (-1.35) (-0.38)
TGi x TPit x RPTit 0.0564* 0.0663** 0.0199
(1.67) (2.23) (1.02)
Constant Yes Yes Yes
Firm FE Yes Yes Yes
Year FE Yes Yes Yes
Observations 610 610 610
Number of firms 72 72 72
within R-sq 0.0972 0.113 0.0975
-
33
Table 8: Falsification Tests on Related-Party Transactions
Firm fixed effects regressions, identical to those reported in Table 4, but using treatment group dummy (TGi),
where non-heirs (controlling shareholder or other relatives) become a major shareholder. Columns (1) – (3) ((4)
– (6)) use 25 (30) treatment group firms that experienced major increase in controlling shareholder’s (other
relative’s) ownership and 25 (30) control group firms identified by covariate matching based on year, industry
(4-digit code for manufacturing and 2-digit code for others), and asset size. t-values are reported in the
parenthesis, and are based on standard errors clustered at the firm level. *, **, and *** respectively indicate
significance at 10%, 5%, and 1% levels. Significant results (at 10% level or better) are shown in boldface.
(1) (2) (3) (4) (5) (6)
Dependent
Variable
Related-party
Transactions
(RPTit)
Related-party
Sales
(RPSit)
Related-party
Purchases
(RPPit)
Related-party
Transactions
(RPTit)
Related-party
Sales
(RPSit)
Related-party
Purchases
(RPPit)
Treatment Group Controlling Shareholder Other Relatives
TPit -0.0445 -0.2236 -0.3127 0.0841 0.3954 -0.1440
(-0.26) (-0.69) (-1.08) (0.46) (1.63) (-0.27)
TGi x TPit 0.2476 0.4820 0.0168 0.8179 0.5802 0.4540
(0.75) (0.85) (0.04) (1.50) (1.23) (0.58)
Firm size 0.5519*** 0.1893 1.0197** 0.5978 0.3947 0.7913
(3.86) (0.76) (2.49) (1.33) (1.04) (1.28)
Firm age 0.6401*** 0.9985* 1.1190* 0.2937 0.9491** -0.5816
(2.69) (1.88) (1.71) (1.08) (2.31) (-0.98)
Leverage 0.0671 -0.9718 3.1549* 3.1539 3.5072 4.6526**
(0.09) (-0.86) (1.93) (1.66) (1.42) (2.41)
Constant Yes Yes Yes Yes Yes Yes
Firm FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Observations 379 379 379 473 473 473
Number of firms 50 50 50 60 60 60
within R-sq 0.186 0.135 0.201 0.408 0.249 0.293
-
34
Table 9: Falsification Tests on Earnings, Dividends, and Group Control
Firm fixed effects regressions, identical to those reported in Tables 5, 6, and 7, but using counterparties of the original treatment group firms (e.g. firms where heir become
a major shareholder) as our new treatment group. TGi takes a value of 1 if firm is the counterparty of the original treatment group firm, and 0 otherwise. TPit takes a value
of 1 if the original treatment group firm of firm experiences a major increase in heir’s net ownership at year t or before. Control variables are suppressed. We use 18
treatment group firms and 18 control group firms identified by covariate matching based on year, industry (4-digit code for manufacturing and 2-digit code for others), and
asset size. t-values are reported in the parenthesis, and are based on standard errors clustered at the firm level. *, **, and *** respectively indicate significance at 10%, 5%,
and 1% levels. Significant results (at 10% level or better) are shown in boldface.
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Dependent Variables EBITDAit DIVit MCIit
Types of RPT
Related
-party
Transactions
(RPTit)
Related
-party
Sales
(RPSit)
Related
-party
Purchases
(RPPit)
Related
-party
Transactions
(RPTit)
Related
-party
Sales
(RPSit)
Related
-party
Purchases
(RPPit)
Related
-Party
Transactions
(RPTit)
Related
-party
Sales
(RPSit)
Related
-party
Purchases
(RPPit)
TPit 4.8967** 2.7009** 3.4178** 2.6272 0.3328 -0.4909 -0.4464 -0.8437 -0.2535
(2.81) (2.31) (2.57) (0.65) (0.11) (-0.15) (-0.29) (-1.23) (-0.25)
RPTit -0.3113 0.0589 0.0333 -0.5571 0.0083 0.0112 -0.0032 0.0226 -0.0586
(-1.63) (0.55) (0.19) (-0.98) (0.03) (0.04) (-0.02) (0.36) (-0.47)
TGi x TPit 0.0891 3.1138 0.3096 -3.9350 0.4450 -2.0819 1.0402 1.3039 0.6710
(0.05) (1.19) (0.19) (-0.69) (0.07) (-0.49) (0.63) (1.52) (0.59)
TGi x RPTit 0.8316* 0.5450 0.4239 1.5417 1.0397 0.5971 -0.0397 -0.0567 0.0506
(1.83) (1.00) (1.31) (1.69) (1.08) (0.91) (-0.23) (-0.76) (0.38)
TPit x RPTit -0.3770** -0.2166** -0.2742** -0.2147 0.0016 0.0510 0.0335 0.0698 0.0180
(-2.75) (-2.32) (-2.57) (-0.74) (0.01) (0.21) (0.28) (1.17) (0.22)
TGi x TPit x RPTit 0.0175 -0.2203 -0.0084 0.5099 0.1869 0.3533 -0.0731 -0.1020 -0.0479
(0.13) (-1.17) (-0.07) (1.15) (0.37) (0.96) (-0.55) (-1.42) (-0.51)
Constant Yes Yes Yes Yes Yes Yes Yes Yes Yes
Control Variables Yes Yes Yes Yes Yes Yes Yes Yes Yes
Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Observations 170 170 170 170 170 170 169 169 169
Number of firms 18 18 18 18 18 18 18 18 18
within R-sq 0.566 0.562 0.560 0.400 0.371 0.380 0.0672 0.135 0.0654
i
i
-
35